Interference effects in implicit and explicit memory.

Abstract

Interference effects are widespread in tests of explicit memory, such as recall and recognition. Implicit memory, in contrast, appears to be highly resistant to such effects. Four experiments were performed to provide a systematic investigation of interference in one implicit memory task, stem completion. In the first two experiments the cues used in the stem completion task uniquely identified one studied item (e.g., test cue ANT______; only one studied item, say, ANTIQUE, began with "ANT"). Performance in a single-list control condition was compared to a three-list interference condition in Experiment 1. No indication of interference, proactive or retroactive, was obtained. In Experiment 2, the effect of similarity of the interfering material to the targets was examined by using neighbours (words that shared all but one letter) in the interfering lists. In addition, completion performance was measured on words from each list. There was no interference attributable to the similarity manipulation, and there was no sign of a buildup of interference across list positions. In the final two experiments, the cues were non-unique, i.e., more than one studied item matched the stem cue (e.g., study ANTIQUE, ANTENNA, ANTLER; test cue ANT______). Because such a test limits the subject to one response out of the three studied alternatives, comparison to a uniquely-cued condition can lead to incorrect conclusions. Instead, an appropriate method of analyzing the data was developed. Using this analysis, proactive interference was detected in the stem completion task. The combination of findings from the unique and non-unique cue conditions suggested that there was no interference during encoding (e.g., no unlearning), but that interference effects arose during retrieval. A search mechanism that could explain the pattern of results was proposed.